Gogolook Improves Scam Detection for 100 Million Users Worldwide Using Real-Time Data and AI

With vishing (voice call–based phishing) widespread around the world, demand is higher than ever for solutions that establish trust for incoming calls. Gogolook provides this trust through products like its flagship B2C app Whoscall, which identifies and filters out spam calls. Whoscall boasts the largest telephone number database in East and Southeast Asia, with 2.6 billion entries.

Since its founding in 2012 in Taiwan, Gogolook has expanded both its product and customer base. From nine international offices across Asia and Europe, it now serves the B2B sector as well as government customers such as the Thailand police force—and is a founding member of the Global Anti-Scam Alliance. Born on AWS, Gogolook began with traditional Amazon Elastic Compute Cloud (Amazon EC2) clusters, then adopted serverless technology, self-hosting Kubernetes clusters on Amazon EC2 instances. After partnering with ScamAdviser in 2024, in a bid to keep overhead low and maximize efficiency of its human resources, Gogolook sought to increase automation and integrate more AWS managed services. 

In 2023, Gogolook adopted Amazon Elastic Kubernetes Service (Amazon EKS), a managed service, to start, run and scale Kubernetes clusters. Then, to streamline database administration, the company migrated from Amazon EC2–hosted MongoDB clusters to Amazon DocumentDB, which is compatible with MongoDB application programming interfaces (APIs). When it came to upgrading its MongoDB database, Gogolook made the strategic decision to offload database management while delivering greater data consistency and scalability. Kalin Chih, tech lead at Gogolook, says, “We love AWS managed services because they reduce our operations and maintenance costs and help automate security checks—making it easier to scale globally and securely as we grow.”

The migration took four months and was achieved in cooperation with AWS Enterprise Support solution architects. The company uses compression enhancements in Amazon DocumentDB to reduce the storage footprint of large datasets, as well as query planner improvements to enhance database read performance. “During the design phase for new projects we consult AWS to provide guidance on how to optimize performance, security, and functionality in a cost-effective way,” says Chih. “I was surprised to hear from our engineers that the migration was much easier than we expected.”

Gogolook’s data pipeline is integrated with Amazon Bedrock, a managed service that provides access to numerous large language models (LLMs) for building generative AI applications. Gogolook uses LLMs on Amazon Bedrock to identify and label calls in Whoscall. Until a few years ago, employees were manually adding labels to numbers that were not auto-labeled from the Gogolook database. This labeling task is now completely automated with AI. The company also uses LLMs to analyze user-uploaded screenshots, social media posts, or advertisements, warning if the content is a scam. Finally, Gogolook has introduced AI for an internal tool integrating the Open WebUI interface with Amazon Bedrock. Developers use the tool to decide on the best foundational model within Amazon Bedrock to fit each use case. 

By migrating to Amazon DocumentDB, Gogolook lowered the TCO for its Whoscall database by 40 percent. Chih explains, “The compression feature in Amazon DocumentDB helps us reduce traffic load while improving database performance, and we can easily upgrade versions with no downtime. Reducing overhead costs has helped us to innovate for our end users. For example, enhancements like our identity security feature help users check if their personal data has been breached due to a malicious event.”

Gogolook has also transformed its labeling workflow with generative AI. By integrating Amazon Bedrock, the company cut labeling time by 16 percent compared to its previous manual process—and the accuracy rate of AI-generated labels matches that of employee-generated ones, at nearly 100 percent. Whoscall now uses generative AI to process millions of user-submitted call reports daily, automatically generating consistent phone number labels across languages—a multilingual task that would have been challenging for employees at Gogolook’s Taipei or Bangkok headquarters. Chih adds, “AI labeling is also much better for scaling because we can instantly add capacity as we grow without hiring and training new team members.”

Despite running its primary database cluster in the AWS Asia Pacific (Tokyo) Region, Gogolook has maintained its low latency targets—even for users in Brazil. The company is currently planning to migrate ScamAdviser to AWS and is evaluating future migrations to the AWS Asia Pacific (Taipei) or Asia Pacific (Bangkok) Region. “We’re confident we can maintain or even improve latency by switching Regions, while ensuring compliance with local data regulations and keeping costs low,” Chih explains.